BraTS21 Tumor Segmentation

This project implements a tumor segmentation model using the DICOM (Brain Tumor) dataset that was converted to NIFTI. It utilizes the SwinUNETR model for segmentation and provides functionality for inference, visualization, and volume calculation.

Features

Tumor segmentation using SwinUNETR model
Sliding window inference for processing large 3D volumes
Visualization of segmentation results and differences between two sets of inputs
Volume calculation for different tumor regions

Models

We provide Swin UNETR models which are pre-trained on BraTS21 dataset as in the following. The folds correspond to the data split in the json file.

Name	Fold	Mean Dice	Feature Size	# params (M)	Download
Swin UNETR	0	88.54	48	62.1	model
Swin UNETR	1	90.59	48	62.1	model
Swin UNETR	2	89.81	48	62.1	model
Swin UNETR	3	89.24	48	62.1	model
Swin UNETR	4	90.35	48	62.1	model

Requirements

Python
PyTorch
MONAI
NumPy
Nibabel
Matplotlib

Converssion Requirements

dicom2nifti pydicom

exact versions are given in requirements.txt file

Installation

orignal repository:

git clone https://github.com/yourusername/brats21-tumor-segmentation.git

Install the required packages:
```
pip install -r requirements.txt
```
Download the pre-trained model and place it in the project root directory as model.pt.

Usage

Prepare your DICOM dataset files. The code expects the following file structure:
```
/content/zip/
├── T1
├── T2
├── Flair
└── tice
```
for comparison we need two sets of inputs like these
Update the input_1 and input_2 variables in the main() function with the paths to your input data.
Run the script:
```
python main.py
```
The script will perform inference on both sets of inputs, visualize the results, and save a comparison image as slice_67_comparison.png
Volume calculations and tumor growth/shrinkage information will be printed to the console.

Customization

Modify the unique_values list to focus on specific tumor regions (1: necrotic and non-enhancing tumor core, 2: peritumoral edema, 4: enhancing tumor).
Change the roi_size, sw_batch_size, and overlap parameters in the main() function to adjust the sliding window inference behavior.

Output

The script generates:

that which Slice has the highest Volume e.g 6.
Console output with volume calculations and tumor growth/shrinkage information.
A PNG image (output_comparison.png) showing the segmentation results, differences, and highlighted regions for the specified slice.

Acknowledgments

This project uses the DICOM Brain Tumor dataset and the SwinUNETR model architecture.

FilesExpand file tree

README.md

Latest commit

History